topic Re: Raster calculator divide creates anomalous data in ArcGIS Spatial Analyst Questions

Raster calculator divide creates anomalous data

PaulLeonard — Thu, 07 Apr 2011 21:55:21 GMT

I have two 32bit float rasters that behave normally with all raster calculator functions except divide. If I divide raster1/raster2 the resulting raster has an anomalous range listed in the TOC. However, it stores the correct ratio value in the created raster which can be determined by using the identify tool. I've included two photos to display this. county_fta/county_ftb = county_divide The first photo shows the displayed range of the data (county_divide) which is incorrect and further calculations use these data instead of the correct ones that are contained within. The second photo shows the identify tool showing a normal data point chosen randomly. Any help greatly appreciated!

Re: Raster calculator divide creates anomalous data

DanPatterson_Retired — Fri, 08 Apr 2011 00:09:22 GMT

have your ruled this out as simply being a symbology issue? have you tried to float one of the rasters in the raster calculator prior to division to ensure floating point division? Have you investigated the histogram for the data?

Re: Raster calculator divide creates anomalous data

Anonymous User — Fri, 08 Apr 2011 15:00:05 GMT

Original User: whuber

The output looks correct to me--there's nothing anomalous about the range, because the denominator includes both negative and positive numbers. That suggests it may contain numbers close to zero. Dividing by them will produce values of large size (potentially above 10^30), both positive and negative.

Re: Raster calculator divide creates anomalous data

PaulLeonard — Fri, 08 Apr 2011 15:00:10 GMT

have your ruled this out as simply being a symbology issue? have you tried to float one of the rasters in the raster calculator prior to division to ensure floating point division? Have you investigated the histogram for the data?

Dan, thank-you for your input. Yes I have ruled out symbology and yes I have tried multiple float operations in the calculator (which yield the same results). The histogram is unfortunately based on the anomalous data. See photo of stats used for classification and histogram. Also, I have tried to recalculate statistics to no avail. Am I simply overlooking something simple? Any thoughts?

Re: Raster calculator divide creates anomalous data

Anonymous User — Fri, 08 Apr 2011 15:19:38 GMT

Original User: pl1021

The output looks correct to me--there's nothing anomalous about the range, because the denominator includes both negative and positive numbers.

Hey Bill. Thanks. Am I wrong in assuming the ranges of dividing both rasters should calculate as following because those two cells are being divided?

For example the two maximum raster values:

51.86/38.263 = 1.355356

Re: Raster calculator divide creates anomalous data

Anonymous User — Sat, 09 Apr 2011 20:24:15 GMT

Original User: whuber

Am I wrong in assuming the ranges of dividing both rasters should calculate as following because those two cells are being divided?

For example the two maximum raster values:

51.86/38.263 = 1.355356

To see that this is not correct, consider two two-cell grids. Grid contains (10,1)--in that order--and grid contains (10, .000001). The quotient of the two grids is (1, 1000000) with maximum value 1000000, even though the quotient of the grid maxima is 10/10 = 1 and both grids contain values between 0 and 10, nothing greater.